Probabilistic forecasting of >10 MeV solar proton events from flare, CME and radio observations of solar cycles 23–25: physical predictors and verification of tree-ensemble classifiers

Abstract We study the probability that a strong solar flare accompanied by a coronal mass ejection (CME) produces a >10 MeV proton event at the Earth. The sample contains 964 flares of GOES class M2.0 and above observed in 1997–2022 with associated CMEs and radio data; 94 of them are parent flares of NOAA-listed proton events (≥10 pfu). The sample was cleaned of duplicate flare records and erroneous flare coordinates, and 36 predictors were computed from soft X-ray, CME, type II/IV radio, background-subtracted microwave and active-region data. Gradient boosting (CatBoost), Random Forest and a six-parameter logistic model were verified by repeated nested cross-validation and by an independent test on 2011–2022. CatBoost reaches an area under the receiver operating characteristic curve (ROC AUC) of 0.935 ± 0.004 and a Brier skill score of 0.33; at the threshold maximising the Heidke skill score it warns 72% of the proton-producing eruptions with a false-alarm ratio of 0.51. The skill is preserved for solar cycles 24–25 (ROC AUC 0.94, true skill statistic 0.72). The information is carried mainly by type II radio bursts, source longitude, soft X-ray fluence and CME speed; longitude acts as a modulating condition with an effective boundary near E20, and the contribution of the CME speed changes sign near 1000 km s⁻¹. The logistic model reaches ROC AUC 0.92. About a third of the persistent false alarms in 2010–2022 were followed by sub-threshold proton enhancements. Heuristic event labels and duplicate records reduce the apparent ROC AUC to 0.83–0.85.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-11606790/v1
Primary Topic
Solar and Space Plasma Dynamics
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Probabilistic forecasting of >10 MeV solar proton events from flare, CME and radio observations of solar cycles 23–25: physical predictors and verification of tree-ensemble classifiers

Vadim Ozheredov, Igor Davydenkov
2 links
Research Square
Solar and Space Plasma Dynamics
preprint

Probabilistic forecasting of >10 MeV solar proton events from flare, CME and radio observations of solar cycles 23–25: physical predictors and verification of tree-ensemble classifiers

Vadim Ozheredov, Igor Davydenkov
preprint en

Abstract

Abstract We study the probability that a strong solar flare accompanied by a coronal mass ejection (CME) produces a >10 MeV proton event at the Earth. The sample contains 964 flares of GOES class M2.0 and above observed in 1997–2022 with associated CMEs and radio data; 94 of them are parent flares of NOAA-listed proton events (≥10 pfu). The sample was cleaned of duplicate flare records and erroneous flare coordinates, and 36 predictors were computed from soft X-ray, CME, type II/IV radio, background-subtracted microwave and active-region data. Gradient boosting (CatBoost), Random Forest and a six-parameter logistic model were verified by repeated nested cross-validation and by an independent test on 2011–2022. CatBoost reaches an area under the receiver operating characteristic curve (ROC AUC) of 0.935 ± 0.004 and a Brier skill score of 0.33; at the threshold maximising the Heidke skill score it warns 72% of the proton-producing eruptions with a false-alarm ratio of 0.51. The skill is preserved for solar cycles 24–25 (ROC AUC 0.94, true skill statistic 0.72). The information is carried mainly by type II radio bursts, source longitude, soft X-ray fluence and CME speed; longitude acts as a modulating condition with an effective boundary near E20, and the contribution of the CME speed changes sign near 1000 km s⁻¹. The logistic model reaches ROC AUC 0.92. About a third of the persistent false alarms in 2010–2022 were followed by sub-threshold proton enhancements. Heuristic event labels and duplicate records reduce the apparent ROC AUC to 0.83–0.85.

Research Square
Space Research Institute (RU)
National Aeronautics and Space Administration, National Oceanic and Atmospheric Administration, Goddard Space Flight Center
Solar and Space Plasma Dynamics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon